LSTM Neural Network Predicts Metal Oxide Sensor Equilibrium
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing gas monitoring systems are often expensive, inaccurate, or unsuitable for long-term monitoring due to high energy consumption, as they require heating metal oxide sensors for minutes to reach equilibrium, making them inefficient and power-hungry.
Innovation Solution
A low-power gas monitoring system using metal oxide sensors paired with a neural network, specifically a Long Short-Term Memory (LSTM) network, which predicts equilibrium state resistance values based on transient resistance measurements after brief heating, reducing energy consumption and enabling wireless, continuous monitoring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the sensor is heated for minutes to reach equilibrium, then the gas concentration measurement is accurate, but the energy consumption is high
Solution Approach 1:
The system performs preliminary heating for only 200 milliseconds to generate a transient response, then uses an LSTM neural network to predict the equilibrium resistance value without actually waiting for equilibrium. This preliminary action provides sufficient data for accurate prediction while avoiding the energy cost of prolonged heating.
Solution Approach 2:
The LSTM neural network creates a computational copy or model of the equilibrium state based on the transient response data. Instead of physically waiting for the sensor to reach equilibrium, the system uses the neural network to simulate and predict what the equilibrium resistance would be, achieving the same measurement goal with minimal energy input.
2Use of energy by moving object
If the heating time is reduced to 200 milliseconds, then the energy consumption is reduced by 99%, but the measurement accuracy must be maintained
Solution Approach 1:
The system replaces the physical/chemical process of waiting for thermal equilibrium with a computational approach using an LSTM neural network. The mechanical/thermal system (heating and waiting) is substituted with an information processing system that analyzes transient response patterns and predicts the final equilibrium state, achieving the same measurement objective with dramatically reduced energy consumption.
Solution Approach 2:
The LSTM neural network acts as an intermediary between the transient sensor response and the predicted equilibrium value. It processes the brief 200-millisecond transient data and translates it into an accurate prediction of the equilibrium state, bridging the gap between short measurement time and high measurement accuracy.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system achieves accurate gas concentration prediction with a 99% reduction in energy usage, allowing for prolonged battery life and continuous monitoring in applications like animal research and industrial settings, with an average error rate of 0.12% and lasting up to 20 years on a single battery.
Implementation Method 1
metal oxide sensors that change their electric resistances in response to surrounding gas concentrations
Implementation Method 2
heating the sensor for minutes
Data Source
AI summary
Systems and methods of applying a prediction model to metal oxide sensors that change their electric resistances in response to surrounding gas concentrations for predicting equilibrium state resistance values of the sensors. The method includes heating a metal oxide sensor to for a predetermined period of time for the metal oxide sensor to interact with a surrounding gas; sampling transient resistance values of the metal oxide sensor to obtain sampled transient resistance values; determining an electrical resistance of the metal oxide sensor in a chemical equilibrium state of the interaction of the metal oxide sensor and the surrounding gas via applying a neural network; and determining a concentration level of the surrounding gas at the chemical equilibrium state by mapping the determined electrical resistance to a corresponding concentration level of the surrounding gas.